- Added progress tracking (incl. acceptance rate), and a plotting tool for it.
- Fixed too much oversampling when manual bloking (#35). Thanks Lukas Hergt (@lukashergt), Vivian Miranda (@vivianmiranda) and Will Handley (@williamjameshandley)
- Fixed ifort compatibility (#39, PR #42). Thanks Lukas Hergt (@lukashergt)
- Fixed: using deprecated Pandas DataFrame method (#40). Thanks Zack Li (@xzackli)
- Added GetDist output for best-fit (
ignore-prior: True
)
- Added
stop_at_error
for likelihoods -- fixes #43. Thanks Lukas Hergt (@lukashergt)
- Fixed
cobaya-cosmo-generator
in for Python 2 (#37, thanks Lukas Hergt, @lukashergt) and high-DPI screens (#41).
- Added fuzzy matching for names of modules and parameters in a few places. Now error messages show possible misspellings.
- Modules can now be nested, e.g.
planck_2018_lowl.TT
andplanck_2018_lowl.EE
asTT.py
andEE.py
under folderlikelihoods/planck_2018_lowl
.
cobaya-citation
deprecated in favour ofcobaya-bib
. In addition to taking.yaml
input files as below, can now take individual module names.cobaya-doc
added to show defaults for particular modules.- Added menu to
cobaya-cosmo-generator
to show defaults for modules.
- Naming conventions for output files changed!
*.updated.yaml
instead of*.full.yaml
for updated info,*.[#].txt
instead of_[#].txt
for chains, etc (seeOutput
section in documentation).
- New, more efficient minimizer: pyBOBYQA.
- Added full suite of Planck 2018 likelihoods.
- Added late-time source Cl's as a cosmological observable (CAMB only, for now)
- Changed capitalisation of some function and requests (deprecation messages and retrocompatibility added)
- Backported some bug fixes.
- Fixed versions of external codes.
- Planck: Fix for calibration parameter being ignored in CMBlike version of lensing likelihood.
- Added
--version
argument forcobaya-run
- Many bug-fixes for corner-cases
- Importance re-weighting, adding derived parameters, etc.
- Now picklable!
- Support for skip and thin
- Multiple evaluations with new
N
option.
- Updated to version 1.16
- Handles speed-blocking optimally, including oversampling (manual blocking also possible).
- Reworked input/output parameters assignment (documented in DEVEL.rst)
- Removed deprecated
gaussian
- Capitalization for observables now enforced! (fixed
H=H(z)
vsh
ambiguity) - CAMB and CLASS: fixed call without observables (just derived parameters)
- Fuzzy-matching suggestions for options of blocks.
- More liberal treatment of external Python objects, since we cannot check if they are the same between runs. So
force_reproducible
not needed any more! (deprecation notice left)
- clik code updated for compatibility with Python 3 and modern gcc versions
- Updated to 1.0 (installing from master branch, considered stable)
- Updated to ...
- Added P(k) interpolator
- Manual parameter speed-blocking.
- Now installable with
cobaya-install polychord --modules [/path]
- Added "citation" tab.
- More permisive resuming.
- Made possible to fix a parameter whose only role is being an argument of a dynamically defined one.
- Likelihoods can be used in dynamical derived parameters as
chi2__[name]
(cosmological application: added automatic consolidated CMB and BAO likelihoods).
- Seeded runs for
evaluate
,mcmc
andpolychord
.
- Small improvements to callback functions.
- Updated to PolyChord 1.15 and using the official GitHub repo.
- Fixed output: now -logposterior is actually that (was chi squared of posterior).
- Interfaced callback functions.
- Added P(k) interpolator as an observable (was already available for CAMB, but not documented)
- Updated to 2.7.1
- Added P(k) interpolator
- Added Y1 release likelihoods (arXiv:1708.01530)
- Updated to 2015 data (arXiv:1810.05216) and renamed to
bicep_keck_2015
.